Download Workshop on group schemes and p-divisible groups: Homework 1. 1

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Étale cohomology wikipedia , lookup

Algebraic variety wikipedia , lookup

Field (mathematics) wikipedia , lookup

Polynomial ring wikipedia , lookup

Deligne–Lusztig theory wikipedia , lookup

Algebraic K-theory wikipedia , lookup

Fundamental theorem of algebra wikipedia , lookup

Tensor product of modules wikipedia , lookup

Covering space wikipedia , lookup

Complexification (Lie group) wikipedia , lookup

Factorization of polynomials over finite fields wikipedia , lookup

Commutative ring wikipedia , lookup

Algebraic number field wikipedia , lookup

Transcript
Workshop on group schemes and p-divisible groups: Homework 1.
1. Let S be a scheme, and G and G0 group schemes over S.
(i) Using Yoneda’s Lemma and group theory, show that the identity section and inversion morphism for G are uniquely determined by the multiplication morphism and that if f : G → G0 is an
S-morphism that respects the multiplication law then f is a morphism of group schemes. Unwind
the proof of Yoneda’s Lemma to derive at least one of these facts by writing down some diagrams
of morphisms (using well-chosen “test objects”).
(ii) Prove that if f : G0 → H is a map of S-groups then ker f is a normal subgroup of G0 that is
moreover closed if H is S-separated.
(iii) If S = Spec(k) for an algebraically closed field k and G and G0 are locally of finite type
and smooth over k with G a closed k-subgroup of G0 then deduce that G is a normal k-subgroup
of G0 if G(k) is a normal subgroup of G0 (k). Give a counterexample if the smoothness condition is
dropped.
2. Let R be a ring, and work below with R-groups.
(i) Prove that there are no non-trivial homomorphisms from Gm to Ga .
(ii) If R is reduced, prove that there are no non-trivial homomorphisms from Ga to Gm .
(iii) If ε ∈ R is nonzero and ε2 = 0, use it to construct a non-trivial homomorphism from Ga to
Gm . If moreover R is an Fp -algebra, construct an automorphism of Ga not given by R× -scaling.
3. Let k be a field of characteristic p > 0.
(i) For all n ≥ 1 prove that αpn and µpn are isomorphic as k-schemes, but not as k-groups.
(ii) If a group scheme G acts on a group scheme H over a base scheme S (via group automorphisms
of H), define the notion of semi-direct product G n H as an S-group. Make a non-commutative
semi-direct product µp n αp over Fp .
(iii) Construct a short exact sequence 0 → αp → αp2 → αp → 0 over Fp and prove that it is not
split even scheme-theoretically (let alone as a semi-direct product) over any extension field.
4. Let k be a field of characteristic p > 0, and let F : Ga → Ga be the k-group map given
functorially by the pth-power map on Ga (X) = Γ(X, OX ) for k-schemes X.
(i) Prove that Endk (Ga ) is a non-commutative (for k 6= Fp ) polynomial ring k{F } with the
P
j
relation F c = cp F . That is, prove that the “additive polynomials” over k are precisely
cj T p .
(ii) Prove ker(F n ) = αpn and ker(F − 1) ' Z/pZ. What is ker(F n − 1)?
(iii) Show that if K is a field of characteristic 0, then EndK (Ga ) consists of scalar multiplications.
ζN
5. Prove “by hand” that the diagram 0 → µN → µN N 0 → µN 0 → 0 is a short exact sequence. Use
Cartier duality to give a second proof by pure thought.
6. Let p be a prime, and let W (X, Y ) = ((X + Y )p − X p − Y p )/p ∈ Z[X, Y ].
(i) Prove that for the Fp -scheme G = Spec(Fp [X, Y ]/(X p , Y p )) with composition law
(x, y) · (x0 , y 0 ) = (x + x0 , y + y 0 + W (x, x0 ))
is a commutative group scheme structure. What is inversion?
(ii) Prove G ' αp∨2 and describe the sequence dual to Exercise 3(iii) via self-duality of αp .
7. (i) Prove that the group functor X 7→ GLn (Γ(X, OX )) on the category of schemes is represented
by the scheme GLn = Spec(Z[tij ][1/ det]) (with 1 ≤ i, j ≤ n). What is its Hopf algebra structure?
(ii) Do the same for SLn , and prove that both GLn and SLn are flat over Spec Z with geometric
fibers that are connected and smooth. (For smoothness of geometric fibers for SLn , find the
dimension of the tangent space at the identity by using the dual numbers.)
1
2
(iii) Write the ring map corresponding to the Z-group map det : GLn → Gm , and use the
irreducibility of det(tij ) over any field (proof?) to deduce that the only group scheme maps from
GLn to Gm over a field are detr for r ∈ Z.
(iv) What is the scheme-theoretic intersection of SLn and the diagonally embedded closed subgroup Gm ,→ GLn ? Do this functorially and algebraically.
8. Let k be a field, k/k an algebraic closure, and G a locally finite type k-group.
(i) Prove that a group scheme is separated if and only if its identity section is a closed immersion
(Hint: identity section is base change of the diagonal); deduce that G is k-separated.
(ii) If Gk is smooth, prove that for any subgroup Γ ⊆ G(k) the Zariski closure of Γ in G is a
closed k-subgroup of G whose formation commutes with extension of the base field.
(iii) If k is perfect, prove that Gred is a (closed) k-subgroup of G. Can you find a counterexample
if k is not perfect? For any field k with char(k) = p > 0, show that the natural semidirect product
G = Gm n αp has Gred = Gm a non-normal k-subgroup of G.
(iv) Prove that a connected k-scheme X that is locally of finite type is geometrically connected
if X(k) is non-empty. (Hint: Use local finiteness of the set of irreducible components to reduce to
the quasi-compact case, and show that K ⊗k Γ(X, OX ) → Γ(XK , OXK ) is an isomorphism for any
K/k in this case. Then chase idempotents and study fibers of the map k 0 ⊗k X → X for finite
separable k 0 /k. Actually, with better technique the locally finite type hypothesis can be dropped;
see EGA IV2 , 4.5.13) Deduce via (iii) over k that if G is connected then it is irreducible and in
fact quasi-compact. (Hint: Use (g1 , g2 ) 7→ (g1 g2 , g2 ) to prove that m : G × G → G is flat, hence
open, so m(U × U ) is open in G for any open U ⊆ G.) Can you build a non-quasi-compact and
non-separated flat group locally of finite type over the spectrum of a discrete valuation ring?
(v) If G0 is the identity component of G, prove that G0 is an open and closed normal subgroup of
G. Can you use Galois descent to define an étale “component group” G/G0 such that G → G/G0
has kernel G0 and is universal for k-group maps from G to étale k-groups? (It may be easier to
just consider the case when G is quasi-compact, in which case G/G0 is a finite k-group.)
9. Let S be a scheme and X an S-scheme.
(i) Prove that for any S-scheme T , the presheaf of sets U 7→ X(U ) = HomS (U, X) on T is a
sheaf for the Zariski topology. Deduce that if X and Y are S-schemes and Xaff and Yaff denote the
resulting functors on the category of affine schemes over S (equipped with S-morphisms), then any
natural transformation Xaff → Yaff arises from a unique map of S-schemes X → Y . Hence, we may
functorially work with schemes as covariant functors on rings (including structure maps to a base
scheme), and in particular we may speak of a functor on rings being representable by a scheme.
(ii) Use rings with non-trivial Picard group to show that the covariant group functor R 7→
GLn (R)/R× does not satisfy the “Zariski sheaf” property in (i) for n > 1, and so it is not representable by a scheme.
(iii) Prove that if a covariant functor F on rings is representable by a scheme, then for any ring
R and finite group G ⊆ Aut(R) the natural map F (RG ) → F (R)G is a bijection. Deduce via
n
the possible non-triviality of k × /k × ' H1 (ks /k, µn ) for fields k with char(k) - n that the functor
R 7→ SLn (R)/µn (R) is not representatable by a scheme for n > 1.
10. The miracle flatness theorem (§23 in Matsumura’s Commutative Ring Theory) says that if
A → B is a local map between local noetherian rings with A regular and B Cohen-Macaulay
(e.g., B regular) then the dimension formula dim B = dim A + dim(B/mA B) implies flatness. (The
converse holds without regularity and CM conditions.)
Prove that if f : G → G0 is a surjective map between locally finite type group schemes over a
field k with Gk and G0k smooth then f is faithfully flat.
3
11. (i) If A → A0 is faithfully flat and M is an A-module, prove that M → M 0 = A0 ⊗A M is an
isomorphism onto the A-submodule of elements m0 ∈ M 0 that satisfy p∗1 (m0 ) = p∗2 (m0 ), where
p∗1 , p∗2 : M 0 ⇒ A0 ⊗A A0 ⊗A M
are determined by a0 ⊗ m 7→ a0 ⊗ 1 ⊗ m, 1 ⊗ a0 ⊗ m. (Hint: First assume A → A0 has a section with
kernel ideal I, and use the resulting decomposition A0 = A ⊕ I as A-modules. Then use faithfully
flat base change by itself to reduce to this case via the diagonal section.) In particular, a0 ∈ A0 lies
in A if and only if a0 ⊗ 1 = 1 ⊗ a0 in A0 ⊗A A0 ; give a counterexample if “faithful”
Q is dropped.
(ii) If {Spec Ai } is a finite open affine covering of Spec A prove that A0 = Ai is faithfully flat
over A and express (i) in terms of gluing for the Zariski topology. Likewise, if K/k
Q is a finite Galois
extension of fields with Galois group G then use the isomorphism K ⊗k K ' g∈G K defined by
a ⊗ b 7→ (g(a)b) to express (i) as the statement (K ⊗k V )G = V for any k-vector space V (with G
acting on K ⊗k V through the left tensor factor).
(iii) Let S be a scheme and f : X 0 → X a faithfully flat and quasi-compact S-map. Let
p1 , p2 : R = X 0 ×X X 0 ⇒ X 0
be the two projections. By working Zariski-locally on X and X 0 (i.e., use the first part of Exercise
9(i)) and using that a finite disjoint union of affines is affine, prove that if Y is an S-scheme then
Y (X) = HomS (X, Y ) naturally injects into Y (X 0 ) and is identified with the subset of elements
with the same image under both maps p∗i : Y (X 0 ) ⇒ Y (R). How does this recover Exercise 9(i) as
a special case? (Suggestion: First take S = Spec Z. Then track S-compatibility by taking Y = S.)
(iv) In the setup of Exercise 10, construct an isomorphism G ×G0 G ' G × ker(f ) as k-schemes,
and deduce that if G → Z is a map of k-schemes that is invariant under translation by ker(f ) then
it uniquely factors through f : G → G0 .
(v) Use (iii) to derive a new proof that Cartier duality carries short exact sequences to short
exact sequences. (Hint: recall that finite – even proper – monomorphisms are closed immersions.)
2
12. Let PGLn = Spec(Z[tij ](det) ) = D+ (det) ⊆ Pn −1 with 0 ≤ i, j ≤ n − 1.
(i) Check that Z[tij ](det) is a Hopf subalgebra of the coordinate ring of GLn , and prove that the
natural map GLn → PGLn is a faithfully flat homomorphism with kernel given by the diagonally
embedded Gm ,→ GLn .
(ii) Functorially identify GLn (R)/R× with a subgroup of PGLn (R), and show that for any
M ∈ PGLn (R) there is a Zariski-open covering {Spec Rri } of Spec R (with ri ∈ R) such that
the image of M in PGLn (Rri )) is in the subgroup GLn (Rri )/Rr×i . Prove also that for n > 1,
GLn (R)/R× = PGLn (R) if Pic(R) is trivial (e.g., R local).
(iii) Formulate a universal property for GLn → PGLn akin to Exercise 11(iv).
(iv) Prove that the composite homomorphism SLn → PGLn is faithfully flat (use Exercise 10
on fibers over Spec Z, and also the fibral flatness criterion), and that its kernel is the diagonally
embedded µn ; also formulate a universal property of this homomorphism (suggesting that one
should say PSLn = PGLn ). Show also that SLn (R)/µn (R) is naturally a subgroup of PGLn (R),
with equality for a local ring R if and only if R× = (R× )n .